Mashr models: Sample Size and Expression Prediction Models

[Dorothy Chen]

Good afternoon MetaXcan team, 

Thank you so much for putting together this wonderful resource for the community. I have a broad, conceptual question about the mashr prediction models described in the predictDB FAQs (https://docs.google.com/spreadsheets/d/1kXD0rT7HAUetlBm-X5ZohP4Jbl31tUA3bAbteR-xv2Q/edit#gid=0) and in Barbeira et al., 2021 (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-020-02252-4). 

More specifically, I was wondering what factors might account for the lower number of whole blood expression models (12,623 from n = 573 white postmortum participants) compared to a tissue matrix like prostate cancer (14,450 from n = 186) that is trained on far fewer participants. Do you have any thoughts on why this might be the case, as well as any resources I can review? 

Thank you and take care. 

[Dorothy Chen]

Hi, I just wanted to follow up on my previous message. Thanks, and please let me know if more information would be helpful. 

[Hae Kyung Im]

Thank you Dorothy for bringing this up. Let me check and get back to you.

Haky

--
You received this message because you are subscribed to the Google Groups "PrediXcan/MetaXcan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to predixcanmetax...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/predixcanmetaxcan/001170b6-3165-4b16-a626-b5911ad4353bn%40googlegroups.com.

[Dorothy Chen]

Thank you! I just wanted to check and see if there have been any updates since the last email.